Evaluating the Efficiency of Bivariate Models in Determining Subsidence Susceptibility of Kashan Plain Aquifer

Land subsidence is one of the most important environmental problems that affects agriculture and urban infrastructure. In order to plan for mitigating the damages caused by land subsidence, it is necessary to identify high-risk and prone areas for this phenomenon specially in different plains of Iran. The main purpose of this research was zoning the land subsidence in Kashan aquifer using Dempster-Schafer, Weight of Evidence, Shannon Entropy, and Frequency Ratio. To conduct this research, 14 factors affecting the occurrence of subsidence including elevation, slope, slope direction, surface curvature, distance from stream, distance from fault, distance from road, lithology, land use, longitudinal and transverse curvature, water table changes, distance from mining and well density in aquifer surface were used. Then, determining 108 subsidence positions and 108 non-subsidence positions at the aquifer level, these points were randomly classified as 30% and 70% as validation data and test data, respectively. To evaluate each of the models, the ROC curve was used. The results showed that among the used methods based on the ROC curve and the amount of AUC, the Frequency Ratio method (training: 0.84 and validation: 0.89) performed best and is introduced as the best method for predicting areas with subsidence sensitivity in studied area. Shannon Entropy, Dempster-Schafer and Weight of Evidence were also ranked next, respectively. The results of this study can be used by provincial managers to reduce the damage caused by land subsidence and to determine areas with subsidence sensitivity.